Demonstrations availability for Table-Top 2 Finger Gripper Tasks

[seungbinjoo]

I'm looking into using demonstrations for Table-Top 2 Finger Gripper Tasks. On the tasks page in the documentation, it says these single-arm tasks have demonstrations available:

"PushCube-v1"
"PickCube-v1"
"PushT-v1"
"StackCube-v1"
"RollBall-v1"
"PullCube-v1"
"PokeCube-v1"
"LiftPegUpright-v1"
"PegInsertionSide-v1"
"PullCubeTool-v1"

However, when using python -m mani_skill.utils.download_demo <env_id> to download demos as instructed in the documentation, for most tasks I only get a folder named rl containing trajectories from presumably from a trained rl policy. I can only find demo trajectories within motionplanning folder for these tasks:

"PegInsertionSide-v1"
"PickCube-v1"
"PushCube-v1"
"PullCubeTool-v1"
"StackCube-v1"

Am I doing something wrong when downloading? Or are demonstrations not available yet?

[seungbinjoo]

Oh I see the motionplanning.sh scripts under data_generation. Are the environments in here the ones with motion planning demonstrations? Namely:

PushCube-v1 PickCube-v1 StackCube-v1 PegInsertionSide-v1 PlugCharger-v1 PullCubeTool-v1 LiftPegUpright-v1 PullCube-v1 StackPyramid-v1

[seungbinjoo]

I've been running the motionplanning.sh scripts but I run into this issue, not sure where it comes from (I'm on the correct version of numpy):

for env_id in LiftPegUpright-v1 PullCube-v1 
do
    python -m mani_skill.examples.motionplanning.panda.run --env-id $env_id \
      --traj-name="trajectory" --only-count-success --save-video -n 1 \
      --shader="rt" # generate sample videos
    mv demos/$env_id/motionplanning/0.mp4 demos/$env_id/motionplanning/sample.mp4
    python -m mani_skill.examples.motionplanning.panda.run --env-id $env_id --traj-name="trajectory" -n 1000 --only-count-success
done

Output:

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mani_skill/examples/motionplanning/panda/run.py", line 12, in <module>
    from mani_skill.examples.motionplanning.panda.solutions import solvePushCube, solvePickCube, solveStackCube, solvePegInsertionSide, solvePlugCharger, solvePullCubeTool, solveLiftPegUpright, solvePullCube, solveDrawTriangle, solveDrawSVG, solvePlaceSphere, solveStackPyramid
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mani_skill/examples/motionplanning/panda/solutions/__init__.py", line 1, in <module>
    from .pick_cube import solve as solvePickCube
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mani_skill/examples/motionplanning/panda/solutions/pick_cube.py", line 5, in <module>
    from mani_skill.examples.motionplanning.panda.motionplanner import \
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mani_skill/examples/motionplanning/panda/motionplanner.py", line 1, in <module>
    import mplib
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mplib/__init__.py", line 3, in <module>
    from mplib.planner import Planner
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/mplib/planner.py", line 7, in <module>
    import toppra as ta
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/toppra/__init__.py", line 16, in <module>
    from . import constraint
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/toppra/constraint/__init__.py", line 83, in <module>
    from .linear_joint_velocity import (
  File "/users/joo/miniconda3/envs/4YP/lib/python3.11/site-packages/toppra/constraint/linear_joint_velocity.py", line 3, in <module>
    from toppra._CythonUtils import (_create_velocity_constraint,
  File "toppra/_CythonUtils.pyx", line 1, in init toppra._CythonUtils
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

[StoneT2000]

Are you using maniskill-nightly? Try reinstalling with that.

We have motion planning based demos for only some tasks. We welcome contributions for more scripted solutions for other tasks. otherwise we sometimes provide RL generated demos as well.

[seungbinjoo]

Yes, I was and am using maniskill-nightly. The script above still run into that issue though

[StoneT2000]

try

    pip install --upgrade numpy

[seungbinjoo]

Still running into same issue with numpy upgraded to numpy-2.3.4
(also it seems maniskill depends on numpy-1.26.4?)